ASRock Rack Unveils M.2 Slot Graphics Card

[HammerOn1024]

1920 X 1080 X 4 (32 bit colour) = 8MB. 16MB would allow for double buffering which is enough to ensure smooth GUI/text performance which is all this thing will do.

Uhmmm.... no. That's 4-bit color not 32-bit as you've written it; 16 actual colors.

 

[wolrah]

I think that could be quite handy to someone building a home server on a consumer grade platform like Ryzen, as high core counts CPU don't come with integrated graphics and server grade motherboards with an IPMI module are more expensive and harder to find.

Exactly what I'd be interested in buying it for, if only it had something better than VGA. There's currently a GeForce 970 in my server for really no good reason other than to provide a video output and I'd love to be able to free up those PCIe lanes for another HBA or high speed NIC.

Unfortunately the server world has that annoying fear of change thing going on and somehow VGA is still a thing.

Uhmmm.... no. That's 4-bit color not 32-bit as you've written it; 16 actual colors.

4 bytes = 32 bits

 

[Keullo-e]

4 bytes = 32 bits

Seems to be correct.

How do I calculate video RAM requirements?

I'm upgrading my video card, but I can't decide between 512 or 1024 MB of memory. 512 is cheaper, and do I need 1024? How do I calculate what's needed? Is it based on resolution, settings, ...?

superuser.com

 

[EarthDog]

Why would scalpers bother with speciality parts that arguably have low demand in the first place?

Yeah, can't say that joke landed... nobody wants these. But hey, cat pic and a double post..always worth the lulz!

 

[newtekie1]

Why bother with HDMI when VGA is a lot cheaper and all of the KVMs and infrastructure in datacenters already use VGA as standard? This is for displaying text and very basic graphics, nothing more.

This. This isn't for consumers, it is for people building servers with Threadripper, Ryzen, and Intel chips without iGPUs. It's main use will be rackmount stuff where a normal GPU might not fit or all the PCI-E slots are already used up for RAID cards. You pop this in, and you got a basic display output. That's all a server needs.

 

[n-ster]

Damn I could actually use a couple of these. To be fair, I could just use it headless but it can alleviate headaches

 

[yourboyvic]

we'll call it removable integrated but not quite dedicated graphics

I registered just to say "I LoL'd"

 

[wolrah]

This. This isn't for consumers, it is for people building servers with Threadripper, Ryzen, and Intel chips without iGPUs. It's main use will be rackmount stuff where a normal GPU might not fit or all the PCI-E slots are already used up for RAID cards. You pop this in, and you got a basic display output. That's all a server needs.

For those of us who'd like to move our systems forward, there's a huge difference between continuing to support a legacy standard and exclusively supporting a legacy standard. There's nothing wrong with having a VGA port because a lot of people still need it, but there's no good reason in all but the smallest systems to not also have some form of digital interface. I'm not asking for DisplayPort-over-USBC or anything like that (though you have to admit, if you were starting from scratch with no legacy concerns it'd be a pretty strong contender with everything in a single small and flexible wire), even just upgrading one generation to a DVI-I port would retain 100% compatibility with VGA infrastructure while adding all DVI and HDMI hardware to the list for the cost of passive adapters.

 

[Mussels]

Not this chip in particular, but this form factor.

agreed with that, M.2 2280 or 221110 could totally be used for IGP level GPU's - it's an interesting concept

 

[CheapMeat]

Love the discussion. IGPU basic I think would be so good for trouble shooting or other things. Heck, I use a GT 710 x1 for that now on my HEDT systems.

 

[InVasMani]

Yeah, can't say that joke landed... nobody wants these. But hey, cat pic and a double post..always worth the lulz!

Cat pics make just about everything more funny it's science cats rule and dogs drool it might be junk science, but it's still science.

This. This isn't for consumers, it is for people building servers with Threadripper, Ryzen, and Intel chips without iGPUs. It's main use will be rackmount stuff where a normal GPU might not fit or all the PCI-E slots are already used up for RAID cards. You pop this in, and you got a basic display output. That's all a server needs.

That's what I had figured and a lot of those rack mount server displays you pull out and prop up are VGA.

I registered just to say "I LoL'd"

10 day TPU penalty for not registering to lulz at cat meme...

 

[Jetster]

Put some serous dedicated memory on that thing so its better than a IGPU

 

[Mussels]

I wonder if an entry level GPU in that form factor would be a lot cheaper than a regular budget card? Entry level GPU's cost way too much for what they are

 

[kayjay010101]

Put some serous dedicated memory on that thing so its better than a IGPU

What's the point? It's added cost to accomodate a market that would never want a product like this anyway.

I wonder if an entry level GPU in that form factor would be a lot cheaper than a regular budget card? Entry level GPU's cost way too much for what they are

You'd hope so, considering the lack of complexity and low BOM. But then again it's enterprise, so AsRock will surely find a way to sell there for $500 a piece.

 

[Jetster]

What's the point? It's added cost to accomodate a market that would never want a product like this anyway.

That's one opinion. I can see a use, if it had more than 16 MB memory

 

[kayjay010101]

That's one opinion. I can see a use, if it had more than 16 MB memory

And what's that?

 

[InVasMani]

I wonder if an entry level GPU in that form factor would be a lot cheaper than a regular budget card? Entry level GPU's cost way too much for what they are

Would be quite something if AMD could utilize a infinity cache based GPU chip in this form factor. Plus there is also 3D NAND, direct storage API, and StoreMi. Finally there is SAM with BAR size and PCIE isn't exactly standing still not to mention you could just as easily take this concept and re-purpose it for PCIE x2/x4 wired M.2 devices to top it all off. The other aspect is you could even crossfire them not that I think it would make a great deal of sense you'd probably as well off not doing so and have each with a single output handling their own displays up to 3 would be pretty trivial given a lot of boards already have that many M.2 slots. I think a few of this to add additional displays while offering some upscale and post process assistance potential to a discrete GPU would be a nice thing to consider as well.

I think even Direct X 12 might make that more usable and useful for that matter the infinity cache would be combined would it not I mean that was suppose to be a perk to mGPU for DX12. I think in the case of a triple display setup with a discrete GPU and 3 of these each of these could handle a display and do a bit of upscale while the discrete GPU renders the entire scene and splits it among the 3 M.2 GPU's to displays like SoftTH. If you had 4 they could easy handle a quadrant. to upscale and post process within a certain threshold. The GPU chips on the M.2 would obviously be small die sizes as well so the yields could be good too. It's defiantly entirely possible to do iGPU upscale of discrete graphics even now and post process should certainly be possible too seeing as the mClassic is capable of it. They would be pretty nifty with a touch of co-processor tech capabilities bundled into each.

AMD could even have 3 of these wired to the discrete GPU's and do away with the additional HDMI/display ports in practice and just add one of these types of devices for additional displays that can also do a bit of upscale and post process. The aspect is if you had a 4K display with 4 inputs that can do PIP/PBP of whatever type and 4 outputs say 3 of these and the discrete graphics itself and all had some upscale/post process they could perform on each output that could be rad. I suppose Intel/Nvidia could certainly do these things mentioned as well for that matter, but infinity cache adds a layer of interest to the whole notion of these compact all in one single display GPU's in regard to performance potential.

 

[RAA!]

Alright friends lets get to work and see how much we can OC this thing. I have to perfect cooler for it already:

 

[EarthDog]

LOL, what is that thing? No thanks on my M.2 with that little tiny screamer fan!!!

 

[CheapMeat]

lmao, I want that cooler. There's probably a few like that on Aliexpress.

 

[btarunr]

This can actually be a possible solution for upgradeable laptop graphics card.

Indeed. Today's high-end Qualcomm Snapdragons already handle 1440p and 4K displays on smartphones, and their 3D graphics are highly capable. Imagine putting them on an M.2 card with POP LPDDR4x memory, and a passive M.2 heatsink. Display comes out from the same PCIe x4 interface, accelerating the iGPU (or via header in the absence of an iGPU). I reckon an M.2-2280 PCB has enough space for a Snapdragon mobile AP (with POP memory), and a bridge chip that interfaces the AP with PCIe 3.0 x4. Host-side software handles the rest.

 

[InVasMani]

Alright friends lets get to work and see how much we can OC this thing. I have to perfect cooler for it already: View attachment 178883

That cooler is pretty awesome I want 4 of those installed on a Quad M.2 PCIE x16 card or two of them mounted on the backplate of my GPU for no real reason.